456 



PROCEEDINGS OF THE AMERICAN ACADEMY. 



From (a) Analysis 1 to 5 



132.891 

 132.876 

 132.877 

 132.876 

 132.878 

 132.869 

 132.878 

 132.877 



132.878 



The " probable error " of this result, obtained according to the method 

 of least squares, is only about ± 0.001, if each average is counted as a 

 single determination. Of course, however, this "probable error " gives 

 no clew as to the presence or absence of a possible constant error. The 

 accuracy of the work and the purity of the silver may be better tested 

 by comparing the weight of silver taken with that of the argentic chloride 

 obtained ; thus 22.93252 grams of silver gave 30.46482 grams of chloride, 

 or 75.275 to 100.00 ; while Stas found the ratio 75.276 to 100.00. This 

 agreement is as close as could be desired ; it shows that no weighable 

 trace of iodine remained in the preparations. 



The means (d) and (h) represent by far the most careful work, and the 

 six experiments included in them give the average 132.877 (±0.0007). 

 This seems to be the most trustworthy value of the atomic weight of 

 caesium as obtained by the analysis of the chloride. 



IV. The Analysis of Potassic Chloride. 



As a further test of the purity of the materials and the accuracy of the 

 method it seemed worth while to make a few similar determinations of 

 the atomic weight of i^otassium. Since this value is already well known 

 from the accurate work of Stas, the present work upon this element was 

 far less searching than the preceding investigation. 



For this purpose pure potassic chloride was prepared as follows. 

 " Chemically pure " potassic chloride was dissolved in pure water, and 

 to this was added a carefully prepared solution of chloroplatinic acid 

 solution. The potassium chlorplatinate was well washed and reduced 

 by ignition. The potassic chloride was dissolved with hot water in a 

 platinum vessel and the solution was allowed to stand for some time in 

 order that any chlorplatinate which had not been reduced might separate 

 out. The solution was then evaporated, and the potassic chloride 

 repeatedly precipitated with hydrochloric acid gas. No better method 



